Contribution
A large human test of whether a father’s severe, time-limited hardship shortens the lives of children conceived after it. The setting gives two natural controls: men captured when prisoners were exchanged quickly, against men captured when exchanges stopped; and, within families, children born before the war. It adds a test of reversibility, by the child’s quarter of birth, and argues for a mechanism by eliminating others.
Key insight
The timing of capture, not the fact of it, separates fathers who starved from fathers who did not. Exchanges stopped in mid-1863 over the treatment of black soldiers and their white officers, and camps filled. Among survivors to 1900, wartime records mention scurvy for 11% of non-POWs, 14% of POWs captured before July 1863, 23% of those captured between July 1863 and July 1864, and 14% of those captured after June 1864 (Data and Methods). Only the sons of the 23% group show the effect, so capture as such, and whatever selected men into it, is controlled.
Assumptions
- Exchange-period POWs are a fair control for no-exchange POWs: similar men, captured at a different time. Ex-POWs did differ from non-POWs (earlier enlistment, more labourers, fewer farmers), so those traits are controlled.
- Quarter of birth indexes maternal nutrition in utero. Second-quarter births are taken to have had inadequate late-gestation nutrition, and fourth-quarter births adequate; the justification is in the SI, following Doblhammer and Vaupel (Discussion, “Mechanisms”).
- Proof by elimination. If socioeconomic, family, selection and maternal channels each fail to explain the result, and the pattern (sons only, by quarter of birth) fits epigenetics, then epigenetics is the best explanation.
- Mortality after 45 is the outcome because the linked data suit it best (SI). Earlier deaths are not studied.
Key results
- Sample (Data and Methods). 2,342 children of 732 no-exchange POWs, 2,416 children of 715 exchange-period POWs, and 15,145 children of 4,920 non-POW veterans, all born after 1866 and surviving to 45: 19,903 children of 6,367 fathers. Fathers from two longitudinal Union Army samples drawn from service and pension records; children linked to death records and censuses.
- Model. Cox proportional hazards for years lived after 45, clustered on family, stratified on enlistment year; father’s POW group interacted with child’s sex.
- Between families (Table 1). Sons of no-exchange POWs:
- Specification 1, controls for birth year and father’s characteristics at enlistment: HR 1.110 (SE 0.045), P = 0.009. Sons of exchange-period POWs: 1.016, P = 0.663. No-exchange against exchange: χ²(1) = 4.06, P = 0.044.
- Adding postwar paternal wealth, occupation, county density, birth order and siblings: 1.104, P = 0.012.
- Sons alive at 21 in 1910, with their own occupation and marriage controlled: 1.115, P = 0.021 (8,427 sons).
- Adding maternal lifespan and maternal grandfather’s wealth: 1.106, P = 0.011.
- In the last two, the difference from exchange-period POWs’ sons is P = 0.062 and 0.077.
- Daughters: every hazard ratio between 0.959 and 0.997, none significant.
- Within families (Table 2; 1,067 sons of 342 fathers, 855 daughters of 275 fathers, war-born children excluded). Sons of no-exchange POWs born after the war against their brothers born before: HR 2.225 (SE 0.932), P = 0.056. No-exchange against exchange: χ²(1) = 4.71, P = 0.030; against non-POW: χ²(1) = 2.20, P = 0.138. Daughters: no effect. The authors ask that the magnitude “be interpreted with caution”.
- Quarter of birth (Table 3; 10,313 sons of 5,615 fathers). Sons of no-exchange POWs, relative to non-POWs’ sons born in the second quarter: first quarter 1.090 (P = 0.198), second 1.204 (P = 0.008), third 1.101 (P = 0.134), fourth 1.013 (P = 0.849). Second against fourth, χ²(1) = 4.28, P = 0.039.
- Channels examined and, by the authors’ account, excluded (Results;
mostly SI Tables S11–S16, not seen).
- Sons’ socioeconomic outcomes and family structure were barely affected; sons of no-exchange POWs were, if anything, less often divorced.
- POW status did not predict the wife’s family wealth or the mother’s longevity.
- No-exchange POWs alive in 1900 themselves had a hazard of 1.14 against non-POWs, so there was “little attenuation” from father to son. Controlling for the father’s lifespan changed the sons’ estimates only modestly.
- Excess deaths among sons were mainly from cerebral haemorrhage and, marginally, cancer; none from suicide or accidents. Cause of death comes only from the few states with online indexes; the text says the SI examines excess mortality “for the 26% of children with this information”.
- Comparisons offered. The effect is comparable to that of low childhood socioeconomic status or urban residence then (SI); the Överkalix hazard ratio for paternal undernutrition, 1.11, is of similar size (Discussion).
Claims
| id | claim | strength | support |
|---|---|---|---|
| C1 | Severe paternal hardship as a POW raised the older-age mortality of sons born afterwards by about 10% | moderate: large sample, two control groups, stable across four specifications; but the contrast with exchange-period POWs weakens to P = .06 to .08 with fuller controls | Table 1 |
| C2 | The effect is not explained by permanent family traits or by traits that let fathers survive the camps | weak to moderate: rests on a within-family subset whose main estimate has P = .056 and whose contrast with non-POW families is not significant | Table 2, Discussion |
| C3 | Daughters are unaffected | moderate: consistent nulls in both designs | Tables 1–2 |
| C4 | Adequate maternal nutrition in late gestation reverses the paternal effect | weak: a quarter-of-birth gradient (P = .039 for second against fourth), with nutrition inferred from season, not measured | Table 3, SI |
| C5 | The mechanism is most consistent with epigenetic, possibly Y-linked, transmission | weak: argued by elimination and analogy to animal studies and Överkalix; no biological measurement | Discussion |
Concepts
- exchange and no-exchange periods: captured before July 1863 or after June 1864, when prisoners were soon exchanged, versus between, when they were held in worsening camps.
- hazard ratio: the ratio of instantaneous death rates, here for years lived after 45. The paper glosses 1.11 as “1.11 times more likely to die”.
- reversibility: an inherited effect undone by a later exposure, here maternal nutrition, as in the animal studies the authors cite.
- proof by elimination: arguing for one mechanism because measured alternatives fail to explain the pattern.
Connections
- The question it is filed for (QUESTION-297). It is one of the question’s two epidemiological pieces on the germline route, with Överkalix (Pembrey et al., LIT-1764), which the paper leans on for its sex-specific, Y-linked reading.
- The womb (Heijmans et al., LIT-1751). The paper cites the Dutch Hunger Winter methylation work, and its quarter-of-birth result puts the womb inside the paternal story: the mother’s late-gestation nutrition is what decides whether the father’s hardship shows.
- Parenting and silence (Danieli, LIT-1701). The paper cites chapters from Danieli’s handbook for the psychological channel, then argues against it: an effect on boys born in one season “suggests that psychological transmission to boys only is an unlikely explanation”.
- Dias and Ressler (LIT-1746). Not cited. Both papers exclude social transmission by design and arrive at a biological route, one in mice with a molecular result, one in men with none.
Bearing on the record
- The owner’s thesis (CLAIM-491). The question’s framing, that on
the biological route no self is needed, fits this paper’s account of
itself. What passes is a mortality risk, not distress, memory or
meaning. These are inferences; the paper does not discuss identity or
the self.
- The outcome is death from stroke and cancer after 45. It is the least self-involving outcome one could choose, and the authors find no excess suicides or accidents, which they take as evidence against a psychological route.
- The effect depends on the son’s season of birth, which no identification with the father or his cause could track.
- It bears on the identity route at one point, and against it. In the Conclusion the authors suggest that psychological transmission may be commoner among children of POWs held by Japan, Korea or Vietnam, or of genocide survivors, because Civil War POWs “returned home to a grateful nation. Their status as heroes may have given their experiences meaning.” That is a hypothesis that a group’s narrative, a shared meaning given to the injury, protects descendants rather than injuring them. It is untested here. If it held, the identity route would be a moderator of harm, not only a carrier of it. That is an inference from one sentence.
- The triviality objection (CLAIM-492). Not addressed.
- QUESTION-297. The paper does not supply what the question asks for, a human study that separates germline from uterine, parenting and cultural routes. It separates socioeconomic and family-fixed channels well. It has no data on parenting, on PTSD (unrecognized then) or on family interaction, and its strongest mechanistic result shows the womb mattering.
- A dated note is proposed for QUESTION-297 (section 3). None on CLAIM-491.
- No THEORY filed. One study whose mechanism is argued, not measured.
- No ML instruction.
Limitations
- Mechanism by elimination only. The authors: “we cannot reject definitively cultural or psychological explanations”, and even if the route is epigenetic “we cannot disentangle the impact of the stress of captivity from that of starvation” (Strengths and Limitations).
- No data on what would test the psychological route: interactions in the family, paternal PTSD, children’s physical or mental health.
- Older-age mortality only, and cause of death sparse.
- The within-family estimate is not significant at .05, and the within-family test against non-POW families is not significant at all.
- Quarter of birth is a proxy for maternal nutrition, not a measure.
- Most robustness checks are in the SI, which was not read.
- Generalization unknown, by the authors’ own statement; Civil War POWs faced “neither torture nor murder”.
- Inconsistencies and typos.
- The introduction gives “4,593 children of 1,407 former POWs and 15,310 children of 4,960 non-POW veterans”; the analytical sample and Table 1 give 4,758 children of 1,447 ex-POWs and 15,145 children of 4,920 non-POWs. The difference is not explained in the main text.
- Results: second-quarter sons of no-exchange POWs were 1.2 times likelier to die than sons “of ex-POWs during the no-exchange period”; it should be the exchange period, as the abstract has it.
- Results: second-quarter sons of no-exchange POWs “were 1.1 times more likely to die than sons of non-POWs born during the fourth quarter”. I could not reconcile this with Table 3 (1.204 against 0.933).
- The exchange period is defined as after June 1864, but exchanges are said to have resumed in December 1864.
- “childrens’ years lived”, “programing”, “geneontology”, “veterans’children”. In the references: Heijmans et al. 2008 given as PNAS volume “104” (it is 105); “adpative” (ref. 19); “Pennsylvannia” and “edisserations” (ref. 43).
Open questions
- Does the effect appear in grandsons, as in Överkalix? A third-generation result would separate germline transmission from the son’s own exposure through his mother.
- Would sons of men from the same units who escaped capture, or adoptees, separate the father’s biology from the household he made?
- Did the heroic meaning of captivity in the North protect sons psychologically, as the authors speculate? A comparison with Confederate veterans held in Union camps, who returned to defeat, would test it.